Market Audit

Marketing Audit Orchestrator

Marketing audit engine for /market audit <url>. Orchestrate a comprehensive analysis by launching 5 specialist subagents in parallel, then synthesize their findings into a single client-ready report. This is the most complex marketing skill — it produces the highest-value deliverable.

When this skill is invoked

The user runs /market audit <url>. Fetch the target URL, classify the business, launch 5 specialist subagents in parallel, aggregate findings into a unified report saved to MARKETING-AUDIT.md, and display a terminal summary.

Required input contract

Before launching subagents:

• Target URL — must be reachable; if not, report and ask to verify.
• Auth state — if pages are gated, note what is accessible and flag gated content for manual review; do not invent the missing analysis.
• Cross-skill inputs — if COMPETITOR-REPORT.md or BRAND-VOICE.md exist in the current directory, load them.
• Site size — for single-page sites, adapt the analysis and flag the limited scope in the executive summary.

Compliance, claims, and audit-integrity constraints

• No fabricated competitive data. Competitor scores, named results, and comparison entries must come from real observation of competitor URLs or be explicitly marked illustrative.
• Revenue impact estimates are scenarios, not promises. Always use conservative / moderate / aggressive ranges with confidence labels (High / Medium / Low). Never imply guaranteed outcomes.
• Industry benchmarks must be cited or labeled as estimates. "Industry benchmark CR is 2–4%" needs a source or a 🟡 estimate tag.
• No dark patterns in recommendations (hidden costs revealed late, manipulative opt-outs, forced continuity, "Yes / No-thanks" guilt copy) — FTC-banned and damaging long-term.
• WCAG 2.2 AA accessibility is non-negotiable — contrast ≥ 4.5:1 body / ≥ 3:1 large text, ≥ 16px body, keyboard nav, alt text, focus indicators.
• Privacy compliance (GDPR / CCPA) — explicit opt-in for marketing, no pre-checked boxes, privacy policy near forms, no PII targeting in copy.
• Subagent independence. Each subagent works from the same brief but produces its own findings — do not let one subagent's output bias another. Synthesize at Phase 3.
• Handle partial data gracefully. If a subagent fails or a page is gated, note the gap — never fabricate.
• Mobile-first. Recommendations must work on the 60%+ of traffic that is mobile.

Three-phase workflow

Phase 1 — Discovery

1. WebFetch homepage + up to 5 key interior pages (pricing, about, product/features, blog, contact).
2. Classify business type (SaaS / E-commerce / Agency / Local / Creator / Marketplace) — classification shapes every subagent's analysis focus.
3. Build the page map (homepage, landing pages, pricing, product, about, blog, contact, legal). Store for all subagents.

Full detection-signal table per business type + page map detail: references/audit-workflow.md.

Phase 2 — Parallel subagents

Launch all 5 subagents simultaneously with the business type, page map, and fetched content.

Full per-subagent evaluation checklists + cross-subagent funnel view: references/channel-and-funnel-audits.md.

Phase 3 — Synthesis

Aggregate subagent outputs into a unified report.

Scoring (decision logic)

Marketing Score = (
    Content_Score      × 0.25 +
    Conversion_Score   × 0.20 +
    SEO_Score          × 0.20 +
    Competitive_Score  × 0.15 +
    Brand_Score        × 0.10 +
    Growth_Score       × 0.10
)

Per-grade aggregator emphasis + diagnostic heuristics for low category scores: references/scoring-and-diagnostics.md.

Recommendation classification

Example recommendations per tier: references/scoring-and-diagnostics.md.

Revenue impact

Revenue Impact = Monthly Traffic × Conversion Rate Improvement × Avg Deal Value
e.g. 10,000 visitors × 0.5% conv lift × $99 ARPU = $4,950 / mo

Always provide conservative / moderate / aggressive ranges. Full framework + reporting templates: references/remediation-and-reporting.md.

Validation gates

Before delivering the report:

• All 5 subagents executed (or failures noted with explicit gaps).
• Composite score calculated with correct weights; sum to 100%.
• Quick Wins (5–10), Strategic (3–7), Long-Term (2–5) classified.
• Every recommendation carries a Confidence tag (High / Medium / Low).
• No competitor data is fabricated or unmarked.
• Accessibility / privacy / dark-pattern compliance checked on every recommendation.
• Mobile-first explicitly considered for UX-affecting recommendations.
• Partial-data gaps disclosed in the executive summary.

Output expectations

• Report file: write the unified audit to MARKETING-AUDIT.md in the current directory using the full template.
• Terminal output: render the score, breakdown bars, top 3 quick wins, top 3 strategic moves, and estimated revenue impact range.

Full MARKETING-AUDIT.md template + terminal output format: references/remediation-and-reporting.md. Competitor comparison table, executive summary, recommendation packaging examples: references/examples.md.

Error handling

• URL unreachable — report and suggest verifying the URL.
• Subagent failure — continue with remaining subagents and note the gap.
• Authenticated content — note what was accessible; recommend manual review.
• Single-page or content-thin site — adapt the analysis; flag limited scope.

Cross-skill integration

• If COMPETITOR-REPORT.md exists, incorporate its findings.
• If BRAND-VOICE.md exists, use it to contextualize content analysis.
• Reference other available analyses in the executive summary.

Reference map